There's a reason some people walk away from GenAI thinking it's overhyped. They asked it a vague question, got a vague answer, and decided it wasn't for them. That's not a tool problem. That's a prompting problem.
The quality of what you get back is almost entirely a function of how well you set up the request. And the good news is that writing better prompts is not some advanced skill that takes months to develop. It takes maybe an afternoon to get meaningfully better at it.
Below are five prompts we come back to constantly, across industries and business sizes. Use them as starting points. Adjust the context to fit your situation.
1. The "think like my customer" prompt
This one is useful anytime you're making a decision that affects how customers experience your business, whether that's a pricing change, a new product, a shift in how you communicate, or anything else.
The prompt: "I run a [type of business] that serves [describe your customer]. I'm thinking about [describe the decision]. Play the role of a skeptical customer and tell me every concern, hesitation, or question this might raise for them. Be honest, not reassuring."
The key word is "skeptical." Without it, most models default to being helpful and optimistic. You want the friction. You want to know where people might balk before they actually do.
2. The meeting prep prompt
If you're walking into a difficult conversation, a sales call, a negotiation, or even a performance review, this prompt helps you think through it before it happens.
The prompt: "I have a meeting with [describe the person and their role] on [describe the topic]. My goal is [state your goal clearly]. What are the three most likely ways this conversation could go sideways, and what's my best move in each scenario?"
You're not looking for a script. You're stress-testing your thinking before you're in the room. The scenarios it surfaces are usually things you already knew somewhere in the back of your mind but hadn't worked through.
3. The "make this clearer" prompt
Writing is hard. Most business writing is either too long, too vague, or written for the person sending it rather than the person reading it. This prompt fixes all three problems at once.
The prompt: "Here is something I wrote: [paste your text]. Rewrite it so that a busy person who doesn't know our business could understand it in 30 seconds. Cut anything that isn't essential. Use plain language. Don't make it sound like it was written by AI."
That last instruction matters. Left to its own devices, most models drift toward corporate-sounding language. Telling it not to do that usually fixes it.
4. The competitive reality check
This one is good to run periodically, especially when you're planning something new or feeling confident about where you stand in the market.
The prompt: "I run a [describe your business]. My main competitors are [list them or describe the competitive set]. What are three things my competitors are likely doing better than me right now, and what's a realistic way to close each gap? Don't sugarcoat it."
Again, the "don't sugarcoat it" instruction is doing a lot of work. GenAI models are trained to be helpful and agreeable. You have to explicitly ask for the uncomfortable version.
5. The first draft prompt
Whether it's a job posting, a proposal, a follow-up email, or a social post, this is the fastest way to get from blank page to something workable.
The prompt: "Write a first draft of [describe what you need]. Here's the context: [explain the situation, who it's for, and what it needs to accomplish]. Tone should be [professional / conversational / direct]. Don't make it longer than it needs to be."
The goal here isn't to get something you'll send as-is. The goal is to eliminate the blank page and give yourself something to react to. Editing is much faster than writing from scratch, and a first draft from a GenAI tool is usually good enough to work from.
One thing to keep in mind
These prompts work because they give the model context, a clear task, and a constraint or two. That structure is what separates useful outputs from generic ones. The more specific you are about your situation, the more useful the response will be.
If you try one of these and it doesn't land the way you expected, the fix is almost always more context, not a different tool.